The Geometry of Concepts: Sparse Autoencoder Feature Structure

Open in new window